Subject: Kludgy, nichey SystemC/C/C++ synthesis forced MENT to dump CatC
> "Effective on August 31, 2011, the High Level Synthesis organization,
> which produces Catapult C, will become part of a new entity with
> Calypto Design Systems."
>
> - from www.deepchip.com/wiretap/110825.html
From: George Harper <gharper=user domain=bluespec got calm>
Hi, John,
What does this say about SystemC/C/C++ high-level synthesis (C-synthesis)
when its market leader, Catapult C, who has probably invested the most
in making it work, is spun out to a start-up?
It's a recognition that C-synthesis is not the future of hardware design
as the EDA community has touted -- but instead it's relegated to being a
point tool, with a niche market.
Why is C-based HLS limited to being a point tool in a niche market?
It's not a good approach to complex, general-purpose hardware IP and
SoC design. When designs fall within its sweet spot (i.e. simple
loop-and-array-based algorithms like IFFTs/FIR Filters/etc), it can have
impressive results.
But, C-synthesis fundamentally struggles, and often fails, when it comes
to more complex, data-dependent algorithms like the new adaptive wireless
algorithms, Reed-Solomon, Viterbi, H.264, compression, encryption; and
especially for memory subsystems/caches/controllers, DMAs, bus protocols,
networking interfaces, network-on-chips, interconnects, arbiters, CPUs,
controllers, etc.
These issues are not a failure of the EDA tools nor due to insufficient
investment in the technology behind them -- instead, they are fundamental
to using SystemC/C/C++ as an abstraction for hardware.
Compelling technology with a deep vein for academic research? No doubt.
And, if you're lucky to have your design space always fall within its
narrow sweet spot, it's perhaps an interesting solution. But, a point
tool that effectively addresses only a niche market cannot provide the
foundation for the future of hardware design.
You can imagine that Wally is under a lot of pressure to be realistic
about the prospects for his various business units these days. And, in
that context, eliminating a considerable, ongoing investment in what
will only be a niche market, is a prudent business decision.
In his recent EEtimes.com blog entry, Ron Wilson framed it well when he
concluded: "By passing Catapult to Calypto but keeping Vista, Mentor may
be saying that it still believes deeply in the future of model-based
system design, but that it has come to see high-level synthesis as a
point tool, not as a foundation stone in the methodology."
- George Harper
Bluespec, Inc. Framingham, MA
|